Active contours driven by local probability distributions

In this paper, we propose a novel local-based signed pressure force (SPF) function, which is defined by the probability distributions in a local region for each point along the evolution curve. We incorporate this novel SPF function into a generalized geometric active contour (GAC) model, forming a new segmentation model - Local Probability Distributions based Active Contour (LPDAC) model, which can deal with intensity inhomogeneity and weak boundaries efficiently. In addition, the spatially varying prior probabilities are introduced to assist the detection of the small scale details. Thus the boundaries detected by the proposed model appear to be more accurate. Experimental results on synthetic and real images demonstrate the superior performance of our model.